Deglycosylation of pathological specimens alters performance of diagnostic PDL1 antibodies.

Immunohistochemical (IHC) predictive quantitation of PDL1 expression is obligatory in many cancer entities with improved response to immune checkpoint inhibition in PDL1-positive subgroups. With recent demonstration of increased positivity rates after enzymatic deglycosylation in breast cancer specimens, a comparative analysis with two different antibodies and extended controls was performed in a cohort of head and neck squamous cell cancer samples (HNSCC).Formalin-fixed paraffin-embedded tissue from HNSCC specimens was used for initial on-slide method optimization based on the PNGase F assay. SDS-PAGE and immunoblotting with the PDL1 antibody 28-8 was performed to evaluate deglycosylation efficiency. A tissue micro array of n = 527 tissue cores of 181 patients with HNSCC was used to determine the effects of deglycosylation on staining pattern and intensity with PDL1 antibodies 28-8 and E1L3N.Successful on-slide deglycosylation with PNGase F was confirmed by immunoblot but varied across replicates. Using E1L3N (intracellular binding domain, most probably not glycosylated), mean signal intensity as well as the fraction of PDL1 positive cells was increased by deglycosylation. Opposite effects were observed with 28-8 (extracellular binding domain, glycosylated).Deglycosylation reduces diagnostic performance of the PDL1 antibody 28-8. In contrast, effects for E1L3N are complex and probably involve reduction of off-target binding leading to specifically improved signal intensity. However, enzymatic deglycosylation adds further variance to IHC.